Bringing New Use Cases and Workloads to the Cloud with ® Graphics Virtualization Technology (Intel® GVT-g)

Introduction From the exponential growth of video on the Internet to desktop virtualization initiatives, media-rich workloads represent a growing share of network traffic. In addition, cloud computing models are incorporating robust media. These usages represent opportunities for businesses to drive new revenue streams and reduce overall costs, but only if the media processing can be managed efficiently. Efficiency in today’s world often implies the use of cloud computing, but media workloads have traditionally been difficult to schedule in the cloud due to an inability to access (GPU) offload capabilities for optimal performance. In response to this challenge, graphics virtualization techniques have evolved to allow a media-optimized workload to run on top of a virtualized environment, such as virtualizing the graphics processor for concurrent use to provide direct, dedicated access to a GPU for a single virtualized workload, or providing a pass-through for direct, shared access to the GPU for a number of virtualized workloads.

Choice in Graphics Virtualization Techniques Intel’s comprehensive portfolio of graphics virtualization Trade-offs in performance, capabilities, and sharing technologies, known as Intel® Graphics Virtualization exist among each of the supported types of graphics Technology (Intel® GVT), encompasses three distinct virtualization, where performance refers to direct GPU graphics virtualization approaches, using -d, -s, and -g acceleration, capability indicates consistency in visual modifiers to differentiate among them. Developers can experience, and sharing denotes multiple VMs. The API select one or more techniques from the Intel® GVT portfolio forwarding approach offered by Intel GVT-s is capable of to best suit their respective solutions and business models. supporting an arbitrary numbers of guests via , but this approach limits the versatility of the GPU in cases where • Intel® Graphics Virtualization Technology -s (Intel® GVT-s): support for media or general-purpose GPU workloads is virtual shared graphics acceleration (multiple virtual needed. In terms of raw performance, Intel GVT-d, with machines (VMs) to one physical GPU), also known as Virtual direct pass-through to the GPU, offers the lowest overhead Shared Graphics Adapter (vSGA) and may be ideally suited for environments with small • Intel® Graphics Virtualization Technology -d (Intel® GVT-d): numbers of GPU-intensive workloads or with individual virtual dedicated graphics acceleration (one VM to one workloads capable of fully consuming all available GPU physical GPU), also known as Virtual Direct Graphics resources. The full GPU virtualization provided by Intel Adapter (vDGA) GVT-g offers a solid balance of performance, capabilities, and sharing, allowing near-native performance and full • Intel® Graphics Virtualization Technology -g (Intel® GVT-g): binary compatibility for up to seven guests per server. virtual graphics processing unit (vGPU) (multiple VMs to one physical GPU)

API Forwarding (GVT-s) Direct Pass-thru (GVT-d) Full GPU Virtualization (GVT-g)

VM VM VM VM VM VM Frontend Frontend Graphics Driver Graphics Driver Graphics Driver DirectX* APIs OpenGL* APIs

Backend Graphics Device APIs Driver Model

Hypervisor Hypervisor Hypervisor

Registers Registers Pros: Registers Pros: Pros: • Performance GPU • Performance GPU • Performance GPU • Sharing Render Engine • Capability Render Engine • Capability Render Engine • Sharing Cons: Graphics Memory Cons: Graphics Memory Graphics Memory • No media/GPGPU • No sharing • Lacks compatibility Harnessing Near-Native Experiences Benefits with Full GPU Virtualization • Provides near-native performance for visual computing Intel GVT-g offers a full GPU virtualization approach tasks inside a VM by allowing direct hardware execution of with mediated pass-through support for Intel® Processor performance-critical operations Graphics. This approach maintains a virtual GPU instance • Delivers Intel Processor Graphics for each VM, with direct assignment of performance- benefits to multiple VMs simultaneously, with zero porting critical resources. Running a native graphics driver inside a VM (without hypervisor intervention in performance- effort needed by the customer to migrate visual computing critical paths) delivers an optimum balance of near-native tasks to the cloud performance, capabilities, and sharing. Armed with the • Supports up to seven virtual GPUs in parallel on 5th and ability to take optimum advantage of Intel Processor 6th generation Intel® Core™ processors with Intel Processor Graphics within VMs, cloud service providers can deliver Graphics as well as Intel® Xeon® processors E3 family with cloud solutions optimized for visual computing that offer Intel Processor Graphics, with more VMs possible in future experiences equivalent to running on bare metal. This generations approach is supported starting with 4th generation Intel® Core™ processors, with a solid GPU virtualization roadmap • Offers total cost of ownership (TCO) savings on Intel Xeon on the horizon. processor E3 family-based platforms with Intel Processor Graphics as well as the full, open-source implementation of In addition, in the spirit of its upstream-first philosophy Intel GVT-g technology as a member of the open-source community, Intel works upstream to ensure that full, open-source implementations of Intel GVT-g technology exist for Usages open-source virtualization hypervisors, Kernel-based • Media decode/encode/transcode in the cloud, enabling Virtual Machine (KVM) and *, known as KVMGT and greater flexibility and improved resource utilization in XenGT respectively. KVMGT and XenGT deliver excellent balancing media workloads without relying on expensive, virtual GPU performance into VMs, delivering > 85 percent single-purpose DSP equipment 3D performance in VMs compared to native.1 These implementations support up to seven virtual GPUs in parallel • Visual understanding workloads, in which traditional on 5th and 6th generation Intel® Core™ processors (Intel® computer vision techniques are augmented with big data Core™ i3, Intel® Core™ i5, and Intel® Core™ i7 processors) with analytics and machine learning Intel Processor Graphics as well as Intel® Xeon® processors • Virtual Desktop Infrastructure (VDI) that can support E3 v4 family with Intel Processor Graphics. remote desktop sharing from thin clients, while running heavy GPU workloads on a server, such as remote desktops Host VM VM VM and workstations for high-demanding workloads, graphics Graphics Graphics Graphics editing, drafting, and CAD software Driver Driver Driver • In-vehicle infotainment (IVI), combining the dashboard and entertainment systems onto a single CPU/GPU to reduce Intel GVT-g cost and complexity • Classroom e-learning, with benefits similar to VDI for license consolidation and optimized resource utilization Hypervisor To learn more about Intel® Graphics Virtualization Technology (Intel® GVT), visit: Registers GPU https://01.org/igvt-g Render Engine http://www.intel.com/visualcloud Graphics Memory

1 Server: Intel BDW-H SDP. Processor type: Intel® Core i7-585HQ @ 2.70 GHz. GPU Model: Iris Pro 6200 GT3 (0x20). Memory: 16GB 1600 MHz. BIOS settings: HT Enabled: Yes. GTT Size: 8MB. Aperture Size: 1024MB. RC6: Enable. C State: Enable. Host Operating system/Kernel: Ubuntu 14.04.3/drm-intel 4.3.0-rc6. Guest Operating system/Kernel: Ubuntu 14.04.3/drm-intel 4.3.0-rc6. Software and workloads used in performance tests may have been optimized for performance only on Intel® microprocessors. Performance tests, such as SYSmark* and MobileMark*, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Results have been estimated or simulated using internal Intel analysis or architecture simulation or modeling, and provided to you for informational purposes. Any differences in your system hardware, software or configuration may affect your actual performance. INFORMATION IN THIS DOCUMENT IS PROVIDED IN CONNECTION WITH INTEL® PRODUCTS. NO LICENSE, EXPRESS OR IMPLIED, BY ESTOPPEL OR OTHERWISE, TO ANY INTELLECTUAL PROPERTY RIGHTS IS GRANTED BY THIS DOCUMENT. EXCEPT AS PROVIDED IN INTEL’S TERMS AND CONDITIONS OF SALE FOR SUCH PRODUCTS, INTEL ASSUMES NO LIABILITY WHATSOEVER, AND INTEL DISCLAIMS ANY EXPRESS OR IMPLIED WARRANTY, RELATING TO SALE AND/OR USE OF INTEL PRODUCTS INCLUDING LIABILITY OR WARRANTIES RELATING TO FITNESS FOR A PARTICULAR PURPOSE, MERCHANTABILITY, OR INFRINGEMENT OF ANY PATENT, COPYRIGHT OR OTHER INTELLECTUAL PROPERTY RIGHT. UNLESS OTHERWISE AGREED IN WRITING BY INTEL, THE INTEL PRODUCTS ARE NOT DESIGNED NOR INTENDED FOR ANY APPLICATION IN WHICH THE FAILURE OF THE INTEL PRODUCT COULD CREATE A SITUATION WHERE PERSONAL INJURY OR DEATH MAY OCCUR. Intel may make changes to specifications and product descriptions at any time, without notice. Designers must not rely on the absence or characteristics of any features or instructions marked “reserved” or “undefined.” Intel reserves these for future definition and shall have no responsibility whatsoever for conflicts or incompatibilities arising from future changes to them. The information here is subject to change without notice. Do not finalize a design with this information. The products described in this document may contain design defects or errors known as errata which may cause the product to deviate from published specifications. Current characterized errata are available on request. Contact your local Intel sales office or your distributor to obtain the latest specifications and before placing your product order. Copies of documents which have an order number and are referenced in this document, or other Intel literature, may be obtained by calling 1-800-548-4725, or by visiting Intel’s Web site at www.intel.com. Copyright © 2016 Intel Corporation. All rights reserved. Intel, the Intel logo, Intel Core, and Xeon are trademarks of Intel Corporation in the United States and other Countries. * Other names and brands may be claimed as the property of others. Printed in USA 0516/NKR/MESH/PDF Please Recycle